A method, system, device, and storage medium for SOC equalization of a battery system.

Abstract

This invention belongs to the field of battery SOC balancing, specifically relating to a battery system SOC balancing method, system, device, and storage medium. By predicting the SOC, balancing strategies from different balancing algorithms are obtained, and initial populations and extraction probabilities are obtained through fitness values. Extraction populations are then obtained from the initial populations, and batteries are classified according to their characteristics to obtain similar batteries. During crossover, batteries of the same type are bound together to obtain a crossover strategy. Dynamic mutation probabilities are set to obtain an optimized balancing strategy. Combined with a new fitness function, the process iterates continuously to obtain the optimal balancing strategy. This application selects and combines the advantages of different balancing strategies from multiple balancing strategies to obtain a more accurate optimized strategy. Compared to the traditional method of setting fixed mutation probabilities, this application considers the different parameters of different batteries and sets dynamic mutation probabilities, resulting in a better balancing strategy and extended battery life.

Description

Technical Field

[0001] This invention belongs to the field of battery SOC balancing, specifically relating to a battery system SOC balancing method, system, device and storage medium. Background Technology

[0002] Lithium-ion battery energy storage is currently the most widely used energy storage technology. To address the issue of inconsistent State of Charge (SOC) that often occurs during the charging and discharging process, experts in the field have proposed SOC prediction models (such as Long Short-Term Memory networks LSTM) and numerous balancing algorithms (such as passive balancing, active balancing, and multi-level balancing).

[0003] The core advantage of the SOC prediction model is its ability to predict the SOC at the next moment, transforming SOC balancing from a passive response to an active prevention. However, the prediction model itself lacks decision-making capabilities, so an balancing algorithm needs to be introduced to formulate balancing strategies based on the prediction results for optimization.

[0004] However, different equilibrium strategies have different focuses. Since conventional equilibrium algorithms do not have optimization capabilities, the generated equilibrium strategy may not be the optimal strategy, resulting in an equilibrium result that differs greatly from the expected value. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a SOC balancing method, system, device and storage medium for a battery system. By combining the advantages of different balancing algorithms, the balancing strategy is further optimized to obtain a more accurate balancing strategy and extend battery life.

[0006] A method for SOC equalization of a battery system, comprising:

[0007] Acquire current and historical operating data of the battery system, including voltage, current, temperature, and SOC. The battery system comprises several batteries.

[0008] Based on the current operating data and historical operating data, predict the SOC of the next operating cycle;

[0009] Based on the current SOC and the SOC of the next operating cycle, several balancing strategies of the battery system are obtained. The balancing strategy consists of the discharge balancing current ratio of several batteries.

[0010] The corresponding balancing strategy is executed on the battery system to obtain the updated SOC of each battery after each balancing strategy is executed. The fitness value of each strategy is calculated based on the updated SOC of the battery. A new fitness function is constructed based on the fitness value, the maximum value, the minimum value and the average value of the fitness.

[0011] The battery characteristics are obtained, and the batteries are classified according to the battery characteristics to obtain batteries of the same type. Based on the fitness value, the discharge equalization current ratio of batteries of the same type is subjected to cross-mutation operation to obtain several optimized equalization strategies and elite individuals of the battery system.

[0012] The elite individuals and several optimal equilibrium strategies will be used as the new initial population for the next iteration;

[0013] Based on the new fitness function and the optimized equilibrium strategy, the fitness value of the optimized equilibrium strategy is obtained. Crossover and mutation operations are then performed on the new initial population until the set number of iterations is reached or the fitness value no longer increases significantly, thus obtaining the optimal equilibrium strategy. The SOC equilibrium of the battery system is then performed using the optimal equilibrium strategy.

[0014] Optionally, the step of acquiring battery characteristics, classifying batteries based on these characteristics to obtain batteries of the same type, and performing a cross-mutation operation on the discharge equalization current ratio of batteries of the same type based on the fitness value, yields several optimized equalization strategies and elite individuals for the battery system, including:

[0015] Select the top n equilibrium strategies with the highest fitness as elite individuals, and use the remaining equilibrium strategies as the initial population.

[0016] Based on the fitness value of each equilibrium strategy, the sampling probability of each equilibrium strategy is calculated.

[0017] Based on the extraction probability, a balanced strategy is extracted from the initial population until the number of extractions equals the preset number, thus obtaining the extracted population.

[0018] Obtain battery characteristics of the battery, including temperature and internal resistance;

[0019] Based on the battery characteristics, the batteries in the extracted population are classified to obtain batteries of the same type;

[0020] Based on the equalization strategy, the discharge equalization current ratio of each battery in the same type is cross-crossed in pairs to obtain the cross-cross strategy.

[0021] The probability of battery variation is obtained based on the battery temperature and SOC.

[0022] Based on the crossover strategy and mutation probability, an optimized balancing strategy for the battery system is obtained.

[0023] Optionally, the step of executing the corresponding balancing strategy on the battery system to obtain the updated SOC of each battery after each balancing strategy is executed, calculating the fitness value of each strategy based on the updated SOC of the battery, and constructing a new fitness function based on the fitness value, the maximum value, the minimum value, and the average value of the fitness value includes:

[0024] The corresponding equalization strategy is applied to the battery system to obtain the updated SOC of each battery, and the SOC range and standard deviation are obtained based on the updated SOC of each battery.

[0025] Obtain the sum of the absolute values ​​of the equalization current of the battery system and the total number of batteries to which a non-zero equalization current has been applied.

[0026] Based on the SOC range, standard deviation, sum of absolute values ​​of battery equalization current, and the total number of batteries with applied non-zero equalization current, the fitness value of each equalization strategy in the battery system is calculated, and a new fitness function is constructed based on the fitness value, the maximum value, the minimum value, and the average value of the fitness.

[0027] Optionally, the step of classifying the batteries in the extracted population according to the battery characteristics to obtain batteries of the same type includes:

[0028] Based on the battery characteristics, a feature vector is established for each battery, and the feature vectors of several batteries are combined into a feature matrix;

[0029] The feature matrix is ​​normalized to obtain a normalized matrix;

[0030] Calculate the Mahalanobis distance between every two batteries in the normalized matrix;

[0031] Set thresholds for the domain radius and the number of core points;

[0032] Based on the aforementioned domain radius, Mahalanobis distance, and core point number threshold, all core points and boundary points are selected.

[0033] Select any core point as the first target core point, obtain all core points within the radius of the target core point's domain, and form a cluster with the target core point. Select any second target core point within the cluster, and obtain all core points within the radius of the second target core point's domain, until there are no more core points within the radius of the domain of all core points within the cluster boundary, thus obtaining a cluster of the same type.

[0034] Add all boundary points within the same cluster to the same cluster to obtain the target cluster;

[0035] Select an outer core point outside the target cluster, obtain all core points within the radius of the outer core point's domain to obtain an outer cluster. If the outer cluster intersects with the target cluster, merge the target cluster and the outer cluster to obtain a new cluster. Continue until all core points and boundary points are assigned to different clusters. Batteries that do not belong to the new cluster or the target cluster are designated as fixed clusters, and batteries in the new cluster or the target cluster are designated as the same type of battery.

[0036] Optionally, the Mahalanobis distance between every two batteries in the normalized matrix is ​​expressed as:

[0037]

[0038] Where D_M(x,y) is the Mahalanobis distance between the two cells, x and y are the cell characteristics of the two cells, and T is the transpose. It is the inverse of the covariance matrix of the feature matrix composed of multiple battery features.

[0039] Optionally, the step of pairwise cross-multiplication of the discharge equalization current ratios of each battery in the same category to obtain the cross-multiplication strategy includes:

[0040] The discharge equalization current ratio of each battery in the same category is bound together to obtain the binding relationship;

[0041] The discharge equalization current ratio corresponding to the cells of the stationary cluster is positionally limited, and the position is limited to remain stationary during crossover.

[0042] Based on the binding relationship and location constraints, the discharge equalization current ratios of the equalization strategies in the extracted population are cross-crossed in pairs to obtain the cross-cross strategy for each battery.

[0043] Optionally, the battery variation probability obtained based on the battery temperature and SOC is expressed as follows:

[0044] ;

[0045] Where P_mutation_i is the mutation probability of the i-th battery, base_rate is the set base mutation probability, α represents the importance of SOC to the mutation rate, β represents the importance of temperature to the mutation rate, SOC_i is the current SOC of the i-th battery, T_i is the temperature of the i-th battery, avg_T is the average temperature, and avg_SOC is the average SOC.

[0046] Optionally, the probability of selecting each battery, calculated based on its fitness value, is expressed as follows:

[0047]

[0048] Where M is the number of batteries, and the fitness of the i-th individual is... , This represents the probability of selection.

[0049] A battery SOC balancing system, comprising:

[0050] The first acquisition module is used to acquire the current operating data and historical operating data of the battery system. The operating data includes voltage, current, temperature and SOC. The battery system includes a number of batteries.

[0051] The prediction module is used to predict the SOC of the next operating cycle based on the current operating data and historical operating data.

[0052] The second acquisition module is used to acquire several balancing strategies of the battery system based on the current SOC and the SOC of the next operating cycle. The balancing strategy is composed of the discharge balancing current ratio of several batteries.

[0053] The first calculation module executes the corresponding equalization strategy on the battery system, obtains the updated SOC of each battery after each equalization strategy is executed, calculates the fitness value of each strategy based on the updated SOC of the battery, and constructs a new fitness function based on the fitness value, the maximum value, the minimum value and the average value of the fitness.

[0054] The mutation module is used to acquire battery characteristics, classify batteries according to battery characteristics to obtain batteries of the same type, and perform cross mutation operation on the discharge equalization current ratio of batteries of the same type according to the fitness value to obtain several optimized equalization strategies and elite individuals of the battery system.

[0055] The elite individuals and several optimal equilibrium strategies will be used as the new initial population for the next iteration;

[0056] The iteration module is used to obtain the fitness value of the optimized equilibrium strategy based on the new fitness function and the optimized equilibrium strategy, and then perform crossover and mutation operations on the new initial population until the set number of iterations is reached or the fitness value no longer increases significantly, thereby obtaining the optimal equilibrium strategy, and performing SOC equilibrium of the battery system with the optimal equilibrium strategy.

[0057] A terminal device includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor loads and executes the computer program, it employs a battery system SOC balancing method.

[0058] A computer-readable storage medium storing a computer program, which, when loaded and executed by a processor, employs a SOC balancing method for a battery system.

[0059] The beneficial effects of this invention are:

[0060] This application first obtains several equilibrium strategies by acquiring the predicted SOC, and then filters them using fitness values ​​to obtain an initial population and extraction probability. Based on the extraction probability, the initial population is selected to obtain an extracted population. Then, based on battery characteristics, the batteries are classified into similar categories. During crossover, batteries of the same category are bound together to obtain a crossover strategy. Considering the different parameters of different batteries, a dynamic mutation probability is set to obtain an optimized equilibrium strategy. This is then iterated using new fitness values ​​until the final optimal equilibrium strategy is obtained. Compared to existing technologies, this application selects and combines the advantages of different equilibrium strategies from multiple strategies to obtain a more accurate optimized strategy. Furthermore, compared to the traditional method of setting a fixed mutation probability, this application considers the different parameters of different batteries and sets a dynamic mutation probability, resulting in a better equilibrium strategy and extending battery life. Attached Figure Description

[0061] Figure 1 This is a schematic flowchart of a battery SOC balancing method according to the present invention. Detailed Implementation

[0062] A battery SOC balancing method, such as Figure 1 As shown, the present invention includes:

[0063] S1. Obtain the current operating data and historical operating data of the battery system. The operating data includes voltage, current, temperature and SOC. The battery system includes several batteries.

[0064] S2. Based on the current operating data and historical operating data, predict the SOC of the next operating cycle;

[0065] Specifically, the existing LSTM model is used for prediction.

[0066] S3. Based on the current SOC and the SOC of the next operating cycle, obtain several balancing strategies for the battery system. The balancing strategy consists of the discharge balancing current ratio corresponding to several batteries.

[0067] Specifically, the equilibrium strategy is obtained mainly through a combination of randomization, heuristics, and warm-start. Unlike the purely random initialization of traditional genetic algorithms, this invention employs a multi-source hybrid initialization strategy, combining random sampling, physical heuristics, and historical best practices to improve the quality of the initial population.

[0068] Of these, randomization accounts for 40% of the total number of strategies.

[0069] The randomized strategy generation method involves randomly generating each gene (corresponding to the equilibrium current ratio of each battery cell) within the range [0, 1] for each strategy vector (individual). This ensures population diversity, explores unknown regions, and covers the solution space as much as possible.

[0070] Heuristics account for 50% of the total number of strategies.

[0071] The current SOC reflects the immediate state of the battery, while the predicted SOC reflects its future state if no balancing measures are taken. The balancing strategy can be adjusted based on the difference between the current SOC and the predicted SOC.

[0072] Strategy 1 is to balance only the battery with the highest current SOC, with a current ratio of 1 for the others and 0 for the others.

[0073] Strategy 2 is to focus on attacking the current highest SOC, specifically by:

[0074] Apply equalizing currents of the same or different proportions to a batch of battery cells with the highest SOC. (A SOC threshold can be set, and current can be applied to cells with an SOC higher than this threshold; this threshold can be considered as the average SOC).

[0075] Strategy 3: Linear SOC, specifically:

[0076] If the current battery SOC is lower than the average SOC by n%, the applied equalization current will decrease by a certain percentage.

[0077] If the SOC is divided, different values ​​of current are applied to the battery in different ranges.

[0078] Alternatively, a combination of the two methods can be used. If the SOC is too low, a fixed slight current can be applied or limited to 0. If the SOC is controllable, the current can be reduced proportionally.

[0079] Strategy 4: A joint ranking based on the current SOC and the predicted SOC, specifically as follows:

[0080] Calculate the weighted sum of the current SOC and the predicted SOC for each cell, then sort them according to this weighted sum, and apply stronger equalization to the top-ranked cells. Different weighting parameters and different equalization currents can be set.

[0081] Strategy 5: A strategy based on SOC change trends, specifically as follows:

[0082] Calculate the difference between the predicted SOC and the current SOC for each battery (ΔSOC = predicted SOC - current SOC). If ΔSOC is positive and large, it indicates that the battery's SOC will increase in the future and requires attention; if it is negative, it may naturally decrease, which can reduce the need for balancing.

[0083] Strategy 6: A dual-objective strategy of current SOC and predicted SOC, specifically as follows:

[0084] Refer to the execution methods of strategies 1 / 2 / 3. One part of the strategy is for batteries with high current SOC, and the other part is for batteries with predicted high SOC.

[0085] Strategy 7: A strategy based on SOC stability, specifically as follows:

[0086] If both the current SOC and the predicted SOC of a battery are high, it needs to be balanced. If the current SOC is high but the predicted SOC is not high, it may not need to be balanced for the time being. If the current SOC is not high but the predicted SOC is high, then preventative balancing is needed.

[0087] The current SOC is not high, but the predicted SOC is high. This means that the SOC of this battery cell will rise to a relatively high level in the future and may become a high SOC cell in the future.

[0088] Preventive equalization: This refers to equalizing individual battery cells before they reach a high state of charge (SOC) to prevent them from becoming truly high-SOC cells in the future, thus avoiding larger SOC differences later. Essentially, it involves applying a relatively high or moderate equalization current.

[0089] Of these, warm start accounts for 10% of the total strategies, and the specific method is as follows:

[0090] Existing active equilibrium algorithms can be used to participate in the formulation of initial candidate equilibrium strategies, a process known as hot start or prior knowledge injection.

[0091] It's important to note that active balancing algorithms are typically control strategies designed for hardware circuits, while our genetic algorithm optimizes the distribution ratio of the discharge balancing current (a continuous value between 0 and 1). Therefore, we need to convert the output of the active balancing algorithm into our strategy representation.

[0092] S4. Execute the corresponding equalization strategy on the battery system to obtain the updated SOC of each battery after each equalization strategy is executed. Calculate the fitness value of each strategy based on the updated SOC of the battery, and construct a new fitness function based on the fitness value, the maximum value, the minimum value and the average value of the fitness.

[0093] Specifically, the new fitness function is used for the next iteration, and the updated SOC of each battery after each balancing strategy is executed can be implemented using an LSTM model.

[0094] The process of executing corresponding balancing strategies on the battery system yields an updated State of Charge (SOC) for each battery after each balancing strategy is executed. The fitness value for each strategy is calculated based on the updated SOC of the batteries, including:

[0095] The corresponding equalization strategy is applied to the battery system to obtain the updated SOC of each battery, and the SOC range and standard deviation are obtained based on the updated SOC of each battery.

[0096] Obtain the sum of the absolute values ​​of the equalization current of the battery system and the total number of batteries to which a non-zero equalization current has been applied.

[0097] The fitness value of each balancing strategy in the battery system is calculated based on the SOC range, standard deviation, sum of absolute values ​​of battery balancing current, and the total number of batteries with non-zero balancing current applied.

[0098] Specifically, the purpose of this step is to calculate the fitness value of each candidate equilibrium strategy (individual) in the current population, and to quantitatively evaluate the merits of each candidate equilibrium strategy. Individuals with higher fitness have a higher probability of being selected in the subsequent process of choosing their parents.

[0099] The fitness function formula is set as follows:

[0100]

[0101] SOC_range: Represents the difference between the maximum and minimum values ​​of all SOCs after implementing this balancing strategy, i.e., the SOC range. The core objective of implementing the balancing strategy is to minimize the differences among the SOCs, hence the introduction of the range. The smaller the range (a range of 0 represents the ideal state where all SOCs are completely consistent), the better the optimization result. 0.01 is introduced as a smoothing term to prevent calculation errors caused by the range being exactly 0.

[0102] SOC_std: Represents the standard deviation of all SOCs after the equalization strategy is implemented. The standard deviation reflects the dispersion of the SOC of each battery; the smaller the standard deviation, the better the optimization result.

[0103] `consume_balance` represents the sum of the absolute values ​​of all balancing currents, reflecting energy loss. A larger value indicates more wasted energy, lower system efficiency, and potentially higher temperature rise. This value represents the cost of the balancing strategy, introduced as a penalty, and is negatively correlated with individual fitness. Its weighting coefficient reflects the importance of energy consumption targets relative to consistency targets, and its value needs to be adjusted based on the actual system (e.g., battery capacity, maximum power of the balancing circuit, heat dissipation capacity). The higher the importance of energy consumption, the larger the weight should be.

[0104] num_active: Represents the total number of batteries with a non-zero balancing current applied. This value should be as small as possible for the following reasons:

[0105] ① Hardware limitations: Equalization circuits (such as switches, resistors, DC-DC modules) may have power or current limits. Equalizing too many circuits at once will exceed the hardware's capabilities.

[0106] ② Thermal Management: The more equalization units that work simultaneously, the greater the total heat generated, and the higher the requirements for the heat dissipation system.

[0107] ③Reliability / Lifespan: Frequent or large-scale simultaneous participation of units in equalization may accelerate the aging of related components (such as switching transistors and resistors).

[0108] ④ Complexity: Reducing the number of units performing equalization simultaneously helps simplify control.

[0109] In addition, to avoid premature convergence and convergence stagnation, it is necessary to reduce the competitiveness of super-intelligent individuals and increase the competitiveness of weaker individuals. This requires processing the fitness function by performing a mapping transformation, such as a linear transformation, on its function range.

[0110] The linear transformation scheme is as follows:

[0111] Prerequisite: The choice of a and b must ensure that the scaled average fitness is equal to the previous average fitness, and the scaled maximum fitness is a specified multiple (greater than 1) of the average fitness.

[0112] That is, the new fitness function:

[0113]

[0114] Where C is the ratio of the expected maximum fitness to the average fitness, f_avg is the average SOC, f_max is the maximum SOC, and f is the calculated fitness value.

[0115] The generated new fitness function can be used as the fitness function for the next generation of evolutionary algorithms.

[0116] S5. Obtain battery characteristics, classify batteries according to battery characteristics to obtain batteries of the same type, and perform cross-mutation operation on the discharge equalization current ratio of batteries of the same type according to the fitness value to obtain several optimized equalization strategies and elite individuals of the battery system.

[0117] The elite individuals and several optimal equilibrium strategies will be used as the new initial population for the next iteration.

[0118] Battery characteristics are acquired, and batteries are classified according to these characteristics to obtain batteries of the same type. Based on the fitness value, a cross-mutation operation is performed on the discharge equalization current ratio of batteries of the same type to obtain several optimized equalization strategies and elite individuals for the battery system, specifically including:

[0119] Select the top n equilibrium strategies with the highest fitness as elite individuals, and use the remaining equilibrium strategies as the initial population.

[0120] Specifically, the most outstanding (highest fitness) elite individuals in the current population are selected and directly participate in the formation of the new generation population without participating in the subsequent process of selecting parents and generating offspring. This avoids the accidental loss of outstanding individuals during the execution of the evolutionary algorithm.

[0121] Based on the fitness value of each equilibrium strategy, the sampling probability of each equilibrium strategy is calculated.

[0122] Based on the fitness value of each equilibrium strategy, the probability of selection for each equilibrium strategy is calculated as follows:

[0123]

[0124] Where M is the number of batteries, and the fitness of the i-th individual is... , This represents the probability of selection.

[0125] Based on the extraction probability, a balanced strategy is extracted from the initial population until the number of extractions equals the preset number, thus obtaining the extracted population.

[0126] Specifically, a roulette wheel selection method is used. A roulette wheel is created based on the probability of each individual being selected, ensuring that individuals with high fitness occupy a large area and have a high probability of being selected. An individual can be selected multiple times until the total number of selected individuals plus the total number of elite individuals equals the initial population size (ensuring population stability).

[0127] Specifically, based on the dynamic determination of system size and computing resources, computational efficiency is maximized while ensuring optimization effects.

[0128] Method 1: Considering the scale of the battery system, the initial population size can be based on the following formula:

[0129]

[0130] Where population_size represents the population size, base_size represents the minimum population size, and multiplier is the number of equilibrium strategies added for each additional battery.

[0131] For example:

[0132] For a small battery system (12 batteries), with a minimum population size of 20, and an additional 2 equilibrium strategies required for each additional battery, the population size = 20 + 2. 12 = 44.

[0133] For a large battery system (96 batteries), with a minimum population size of 20, and an additional 2 equilibrium strategies required for each additional battery, the population size = 20 + 2. 96 = 212.

[0134] Method 2: Considering computational resources, if a fitness assessment takes t seconds, then the computation time for one generation is approximately population_size × t seconds. If the computation time for each generation does not exceed T seconds, then the maximum population size is T / t.

[0135] The fitness evaluation time for each strategy is related to the number of batteries, for example:

[0136] .

[0137] If a balancer strategy uses m bytes of memory, then the total memory used is population_size × m bytes. If the maximum memory is M bytes, then the maximum population size is M / m.

[0138] The memory usage of each strategy is related to the number of batteries, for example:

[0139] .

[0140] We take smaller sizes from T / t and M / m, and also select smaller populations from Method 1 and Method 2.

[0141] During the iteration process, if slow convergence is observed, the population size can be appropriately increased; if convergence is too fast, the population size can be appropriately decreased.

[0142] Obtain battery characteristics of the battery, including temperature and internal resistance;

[0143] Specifically, the real-time temperature (provided by the BMS temperature sensor) and real-time internal resistance of each battery can be estimated online via electrochemical impedance spectroscopy or excitation response method.

[0144] Each battery is considered a data point, and its two features (temperature and internal resistance) form a two-dimensional vector. All batteries constitute a feature matrix.

[0145] For each battery: X_i = [T_i, R_i], where T_i is the temperature and R_i is the internal resistance.

[0146] Based on the battery characteristics, the batteries in the extracted population are classified to obtain batteries of the same type;

[0147] The process of classifying batteries in the extracted population based on the battery characteristics yields batteries of the same type, including:

[0148] Based on the battery characteristics, a feature vector is established for each battery, and the feature vectors of several batteries are combined into a feature matrix;

[0149] The feature matrix is ​​normalized to obtain a normalized matrix;

[0150] Specifically, Z-Score standardization is used to transform each feature into a distribution with a mean of 0 and a standard deviation of 1.

[0151]

[0152] Where μ is the mean, σ is the standard deviation, T_i is the temperature of the i-th battery, R_i is the internal resistance of the i-th battery, σ_T is the standard deviation of temperature, and σ_R is the standard deviation of internal resistance.

[0153] Calculate the Mahalanobis distance between every two batteries in the normalized matrix;

[0154] The Mahalanobis distance between every two batteries in the calculation of the normalized matrix is ​​expressed as:

[0155]

[0156] Where D_M(x, y) is the Mahalanobis distance between the two cells, x and y are the cell characteristics of the two cells, and T is the transpose. It is the inverse of the covariance matrix of the feature matrix composed of multiple battery features.

[0157] Set thresholds for the domain radius and the number of core points;

[0158] Based on the aforementioned domain radius, Mahalanobis distance, and core point number threshold, all core points and boundary points are selected.

[0159] Select any core point as the first target core point, obtain all core points within the radius of the target core point's domain, and form a cluster with the target core point. Within the cluster, arbitrarily select a second target core point, and obtain all core points within the radius of the second target core point's domain. Repeat this process (arbitrarily select a second target core point within the cluster, and obtain all core points within the radius of the second target core point's domain) until there are no more core points within the radius of the domain of all core points within the cluster boundary, thus obtaining a cluster of the same type.

[0160] Add all boundary points within the same cluster to the same cluster to obtain the target cluster;

[0161] Select an outer core point outside the target cluster, obtain all core points within the radius of the outer core point's domain to obtain an outer cluster. If the outer cluster intersects with the target cluster, merge the target cluster and the outer cluster to obtain a new cluster. Continue until all core points and boundary points are assigned to different clusters. Batteries that do not belong to the new cluster or the target cluster are designated as fixed clusters, and batteries in the new cluster or the target cluster are designated as the same type of battery.

[0162] Specifically, Mahalanobis distance is used. Mahalanobis distance adjusts the correlation between variables through the covariance matrix and is suitable for correlated data. It can amplify the distance between points that are rare in the overall distribution. For example, the distance between a "high temperature, high internal resistance" battery (which may be severely aged) and a "low temperature, low internal resistance" battery (which is relatively healthy) will be calculated to be very large, making it easy to distinguish them into different cooperative groups (similar batteries).

[0163] DBSCAN is used for clustering. DBSCAN is a density-based clustering algorithm that does not require a preset number of clusters (it is unclear how many cooperative groups all batteries will be divided into) and can identify noise points (it can detect anomalies and handle them specially).

[0164] Point classification:

[0165] Core point: If a point's ε-neighborhood contains ≥MinPts points (including itself), then it is a core point.

[0166] Boundary points: Non-core points but located within the ε-neighborhood of core points.

[0167] Noise point: A point that is neither a core point nor a neighbor of any core point.

[0168] Set key parameters:

[0169] MinPts: The minimum number of neighborhood points to form the core point. It is usually taken as the data dimension + 1. For this problem, there are only two dimensions: temperature and internal resistance, so we take 3.

[0170] ε (eps): Neighborhood radius. It is selected using the k-distance method. The distance from each point to the k-th nearest neighbor (the value of k is generally consistent with the value of MinPts) is calculated, sorted from smallest to largest, and plotted. The distance at the "inflection point" (the position where the distance value suddenly increases) in the graph is selected as ε.

[0171] Cluster formation:

[0172] Select all core points. Choose one core point and add all core points within its neighborhood to the cluster. Then, select another core point from all points in the cluster and repeat the above steps until there are no more core points within the neighborhoods of all core points at the cluster boundary. Finally, add all boundary points within the neighborhoods of all core points at the cluster boundary to the cluster.

[0173] Select a core point outside the current cluster and repeat the above steps to form a new cluster. If the new cluster intersects with the previous cluster, then the two clusters are merged into one cluster.

[0174] The remaining points are considered noise points until all core and boundary points are assigned to different clusters.

[0175] The discharge equalization current ratio of each battery in the same category is cross-crossed in pairs to obtain the cross-cross strategy.

[0176] The method of pairwise cross-multiplication of the discharge equalization current ratio of each battery in the same category to obtain the cross-multiplication strategy includes:

[0177] The discharge equalization current ratio of each battery in the same category is bound together to obtain the binding relationship;

[0178] The discharge equalization current ratio corresponding to the cells of the stationary cluster is positionally limited, and the position is limited to remain stationary during crossover.

[0179] Based on the binding relationship and location constraints, the discharge equalization current ratios of the equalization strategies in the extracted population are cross-crossed in pairs to obtain the cross-cross strategy for each battery.

[0180] Specifically, the binding relationship is to bundle strategies of the same group class together and change them as a whole. When crossing, they are crossed as a whole. For example, if there are 5 batteries A, B, C, D, E, and F, where AE is one group, BCD is another group, then when crossing, AE crosses together, BCD crosses with another group, and F remains unchanged and does not cross.

[0181] This crossover method helps preserve the integrity of the parent gene blocks and retains the relevant strategy combinations during the balancing process. This is a crossover method based on thermodynamic constraints. For example, as mentioned earlier, adjacent batteries with similar temperatures indicate that their internal resistances are close, therefore the effects of their balancing strategies are also similar. Their balancing strategies should be implemented in tandem to achieve co-evolution. The essence of SOC imbalance lies in the differences in internal resistances, resulting in varying heat generation and different charge / discharge efficiencies. Therefore, adjacent batteries with similar internal resistances should adopt coupled balancing strategies to prevent the difference in internal resistance from exacerbating SOC imbalance.

[0182] The probability of battery variation is obtained based on the battery temperature and SOC.

[0183] The probability of battery variation, derived from battery temperature and SOC, is expressed as follows:

[0184] ;

[0185] Where P_mutation_i is the mutation probability of the i-th battery, base_rate is the set base mutation probability, α represents the importance of SOC to the mutation rate, β represents the importance of temperature to the mutation rate, SOC_i is the SOC of the i-th battery, T_i is the temperature of the i-th battery, avg_T is the average temperature, and avg_SOC is the average SOC.

[0186] Based on the mutation strategy and mutation probability, an optimized balancing strategy for each battery is obtained;

[0187] Specifically, after crossover, a mutation operation is performed, causing genetic mutations that randomly alter certain gene loci in offspring individuals with a very small probability. This helps maintain population diversity and allows the algorithm to escape local optima and explore regions of the search space not covered by the parents. However, mutation is not always necessary.

[0188] S6. Based on the new fitness function and the optimized equilibrium strategy, obtain the fitness value of the optimized equilibrium strategy, and perform crossover and mutation operations on the new initial population until the set number of iterations is reached or the fitness value no longer increases significantly, thereby obtaining the optimal equilibrium strategy. Perform SOC balancing of the battery system using the optimal equilibrium strategy.

[0189] A battery SOC balancing system, comprising:

[0190] The first acquisition module is used to acquire the current operating data and historical operating data of the battery system. The operating data includes voltage, current, temperature and SOC. The battery system includes a number of batteries.

[0191] The prediction module is used to predict the SOC of the next operating cycle based on the current operating data and historical operating data.

[0192] The second acquisition module is used to acquire several balancing strategies of the battery system based on the current SOC and the SOC of the next operating cycle. The balancing strategy is composed of the discharge balancing current ratio of several batteries.

[0193] The first calculation module executes the corresponding equalization strategy on the battery system, obtains the updated SOC of each battery after each equalization strategy is executed, calculates the fitness value of each strategy based on the updated SOC of the battery, and constructs a new fitness function based on the fitness value, the maximum value, the minimum value and the average value of the fitness.

[0194] The mutation module is used to acquire battery characteristics, classify batteries according to battery characteristics to obtain batteries of the same type, and perform cross mutation operation on the discharge equalization current ratio of batteries of the same type according to the fitness value to obtain several optimized equalization strategies and elite individuals of the battery system.

[0195] The elite individuals and several optimal equilibrium strategies will be used as the new initial population for the next iteration;

[0196] The iteration module is used to obtain the fitness value of the optimized equilibrium strategy based on the new fitness function and the optimized equilibrium strategy, and then perform crossover and mutation operations on the new initial population until the set number of iterations is reached or the fitness value no longer increases significantly, thereby obtaining the optimal equilibrium strategy, and performing SOC equilibrium of the battery system with the optimal equilibrium strategy.

[0197] This application also discloses a terminal device, including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor loads and executes the computer program, it employs a battery system SOC balancing method.

[0198] The terminal device can be a computer device such as a desktop computer, a laptop computer, or a cloud server. The terminal device includes, but is not limited to, a processor and a memory. For example, the terminal device may also include input / output devices, network access devices, and buses.

[0199] The processor can be a central processing unit (CPU). Of course, depending on the actual use, it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc., and this application does not limit it.

[0200] The memory can be an internal storage unit of the terminal device, such as a hard disk or RAM of the terminal device, or an external storage device of the terminal device, such as a plug-in hard disk, smart memory card (SMC), secure digital card (SD), or flash memory card (FC) equipped on the terminal device. Furthermore, the memory can be a combination of an internal storage unit and an external storage device of the terminal device. The memory is used to store computer programs and other programs and data required by the terminal device. The memory can also be used to temporarily store data that has been output or will be output. This application does not limit this.

[0201] In this terminal device, the SOC balancing method of a battery system in the above embodiment is stored in the memory of the terminal device and loaded and executed on the processor of the terminal device for convenient use.

[0202] This application also discloses a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it employs a SOC balancing method for a battery system as described in the above embodiments.

[0203] The computer program can be stored in a computer-readable medium. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain middleware. The computer-readable medium includes any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the computer-readable medium includes, but is not limited to, the above-mentioned components.

[0204] The SOC balancing method of a battery system in the above embodiments is stored in the computer-readable storage medium and loaded and executed on the processor to facilitate the storage and application of the above method.

[0205] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of protection of this application is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this application as described above, which are not provided in detail for the sake of brevity.

[0206] One or more embodiments in this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments in this application should be included within the protection scope of this application.

Claims

1. A method of SOC equalization of a battery system, characterized by, The method comprises the following steps: obtaining current operation data and historical operation data of a battery system, the operation data comprising voltage, current, temperature and SOC, and the battery system comprising a plurality of batteries; predicting the SOC of the next operation cycle according to the current operation data and the historical operation data; obtaining a plurality of balancing strategies of the battery system according to the current SOC and the SOC of the next operation cycle, the balancing strategies being composed of discharging balancing current ratios corresponding to the plurality of batteries; executing the corresponding balancing strategies on the battery system to obtain the updated SOC of each battery after the execution of each balancing strategy, calculating the fitness value of each strategy according to the updated SOC of the battery, and constructing a new fitness function according to the fitness value, the maximum value, the minimum value and the average value of the fitness value; obtaining battery characteristics, classifying the batteries according to the battery characteristics to obtain batteries of the same type, performing crossover and mutation operations on the discharging balancing current ratios of the batteries of the same type according to the fitness value to obtain a plurality of optimized balancing strategies and elite individuals of the battery system; taking the elite individuals and the plurality of optimized balancing strategies as new initial populations for the next iteration; obtaining the fitness value of the optimized balancing strategies according to the new fitness function and the optimized balancing strategies, and performing crossover and mutation operations on the new initial populations again until a set number of iterations is reached or the fitness value no longer significantly increases, to obtain an optimal balancing strategy, and executing the SOC balancing of the battery system by using the optimal balancing strategy.

2. The method of claim 1 wherein, The method of obtaining battery characteristics, classifying the batteries according to the battery characteristics to obtain batteries of the same type, performing crossover and mutation operations on the discharging balancing current ratios of the batteries of the same type according to the fitness value to obtain a plurality of optimized balancing strategies and elite individuals of the battery system comprises the following steps: selecting the first n balancing strategies with the highest fitness value as elite individuals, and taking the remaining balancing strategies as initial populations; calculating the extraction probability of each balancing strategy according to the fitness value of each balancing strategy; extracting balancing strategies from the initial populations according to the extraction probability until the number of extracted balancing strategies is equal to a preset number, to obtain an extracted population; obtaining battery characteristics of the batteries, the battery characteristics comprising temperature and internal resistance; classifying the batteries in the extracted population according to the battery characteristics to obtain batteries of the same type; performing crossover on the discharging balancing current ratios of each battery in the batteries of the same type according to the balancing strategies to obtain crossover strategies; obtaining the mutation probability of the batteries according to the temperature and the SOC of the batteries; obtaining the optimized balancing strategies of the battery system according to the crossover strategies and the mutation probability.

3. The method of claim 1 wherein, The method of classifying the batteries in the extracted population according to the battery characteristics to obtain batteries of the same type comprises the following steps: establishing a feature vector of each battery according to the battery characteristics, and establishing a feature matrix of the feature vectors of the plurality of batteries; normalizing the feature matrix to obtain a normalized matrix; calculating the Mahalanobis distance between each two batteries in the normalized matrix; setting a field radius and a core point number threshold; selecting all core points and boundary points according to the field radius, the Mahalanobis distance and the core point number threshold; Selecting an arbitrary core point as a first target core point, obtaining all core points within a domain radius of the target core point, and clustering the target core point, selecting an arbitrary second target core point in the cluster, obtaining all core points within a domain radius of the second target core point, and obtaining a same-type cluster until there is no core point within the domain radius of all core points in the cluster boundary; Adding all boundary points in the same-type cluster to the same-type cluster to obtain a target cluster; Selecting an outer core point outside the target cluster, obtaining all core points within a domain radius of the outer core point, obtaining an outer cluster, and merging the target cluster and the outer cluster to obtain a new cluster if the outer cluster intersects the target cluster, until all core points and boundary points are divided into different clusters, taking a battery not belonging to the new cluster or the target cluster as a static cluster, and taking a battery in the new cluster or the target cluster as a same-type battery.

4. The method of claim 3 wherein, The calculation of the Mahalanobis distance between each two batteries in the normalized matrix is represented as: where D_M(x, y) is the Mahalanobis distance between two batteries, x and y are the battery features of two batteries, T is the transpose, is the inverse of the covariance matrix of the feature matrix composed of the plurality of battery features.

5. The method of claim 2 wherein, The crossing strategy obtained by crossing the discharge equalization current proportion of each battery in the same-type battery pair by pair includes: Binding the discharge equalization current proportion of each battery in the same-type battery to obtain a binding relationship; Position limiting the discharge equalization current proportion of the battery in the static cluster, and the position limiting is to keep static when crossing; According to the binding relationship and the position limiting, the discharge equalization current proportion of the equalization strategy in the extracted population is crossed by pair to obtain the crossing strategy of each battery.

6. The method of claim 1 wherein, The variation probability of the battery is obtained according to the temperature and SOC of the battery, and is represented as: ； Wherein, P_mutation_i is the variation probability of the i th battery, base_rate is the set basic variation probability, a represents the importance of SOC on the variation rate, b represents the importance of temperature on the variation rate, SOC_i is the current SOC of the i th battery, T_i is the temperature of the i th battery, avg_T is the average temperature, and avg_SOC is the average SOC.

7. The method of claim 2 wherein, The extraction probability of each equalization strategy is calculated according to the fitness value of each equalization strategy, and is represented as: where M is the number of batteries, the fitness of the ith individual is , is the extraction probability.

8. A SOC equalization system of a battery system, characterized by, It includes: The first acquisition module is used for acquiring current running data and historical running data of a battery system, and the running data includes voltage, current, temperature and SOC, and the battery system includes a plurality of batteries; The prediction module is used for predicting the SOC of the next running period according to the current running data and the historical running data; The second acquisition module is used for acquiring a plurality of equalization strategies of the battery system according to the current SOC and the SOC of the next running period, and the equalization strategy is composed of a plurality of discharge equalization current proportions corresponding to the batteries; The first calculation module executes the corresponding equalization strategy on the battery system to obtain the updated SOC of each battery after the execution of each equalization strategy, calculates the fitness value of each strategy according to the updated SOC of the battery, and constructs a new fitness function according to the fitness value, the maximum value, the minimum value and the average value of the fitness. A variation module is configured to acquire battery characteristics, classify batteries according to the battery characteristics, obtain same-type batteries, and perform cross variation operation on discharge equalization current ratios of the same-type batteries according to the fitness value to obtain a plurality of optimized equalization strategies and elite individuals of the battery system; The elite individuals and the plurality of optimized equalization strategies are used as new initial populations for next iteration; An iteration module is configured to obtain fitness values of the optimized equalization strategies according to the new fitness function and the optimized equalization strategies, perform cross variation operation on the new initial populations again until a set number of iterations is reached or the fitness value no longer increases significantly, and obtain an optimal equalization strategy to perform SOC equalization of the battery system according to the optimal equalization strategy.

9. A terminal device comprising a memory and a processor, characterized in that, The memory stores a computer program capable of running on the processor, and the processor loads and executes the computer program to adopt the method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored therein a computer program, characterized in that, The computer program is loaded and executed by the processor to adopt the method in any one of claims 1 to 7.

Images